Monitoring Hydraulic Fracture Volume Using Surface to Borehole EM and Conductive Proppant

Abstract

Numerical forward and inverse simulations using measured electric and magnetic field noise spectra and actual deviated well casing geometries are performed. The simulations demonstrate that hydraulic fracturing (HF) using typical slurry volumes of electrically conductive proppant at depths of 3km will produce measurable responses. The measured responses for sequential frac-stages can be inverted with a parametric box model to predict the stimulated volume to within 20%. This study presents the results of forward and inverse simulations of a North American drill pad with six deviated cased wells where the true well trajectory with multiple casing strings are considered. This more accurate representation of multiple complex geometry wells is accomplished using OcTree finite difference modelling in the time domain (Haber & Heldmann, 2007) with conductivity and magnetic susceptibility upscaling in conjunction with data differencing of time lapse measurements. The simulation represents a series of 10 sequential HF stages where electrically conductive proppant is injected into a layered background model derived from inductions logs at the site. Inversion of surface-to-borehole EM data for parameters of an electrically conductive box used to represents a HF zone is possible under the conditions studied here. The strong effect of steel infrastructure such as the well casings require that any numerical scheme used to model HF EM data is capable of accurately modelling the casings near the HF. Modelling that approximates only one or two wells will have errors that exceed the response of the HF, thus rendering accurate inversions unlikely. The volume of the HF zone is the best determined parameter which is important for helping to guide well placement and frac stage separation. In practice, there are significant challenges to accurately characterize all the steel in an HF environment and to increasing our ability to characterize HF geometries and conductivity distributions with greater complexity than a single box. The work to date is however encouraging, showing that there is information in the data and that our numerical modelling capabilities are progressing rapidly. The OcTree codes used here can model all the necessary wells in a parametric inversion. In addition, developments in unstructured finite-element codes is progressing rapidly with the prospect of having a choice between OcTree and finite-element based inversion in the near term.

AAPG Datapages/Search and Discovery Article #90332 © 2018 AAPG International Conference and Exhibition, Cape Town, South Africa, November 4-11, 2018